Learning Competitive and Discriminative Reconstructions for Anomaly Detection
Authors: Kai Tian, Shuigeng Zhou, Jianping Fan, Jihong Guan5167-5174
AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Empirical studies on 7 datasets including KDD99, MNIST, Caltech-256, and Image Net etc. show that our model outperforms the state-of-the-art methods. |
| Researcher Affiliation | Academia | 1Shanghai Key Lab of Intelligent Information Processing, and School of Computer Science, Fudan University, China 2Department of Computer Science, University of North Carolina at Charlotte, Charlotte, NC 28223 USA 3Department of Computer Science & Technology, Tongji University, China |
| Pseudocode | Yes | Algorithm 1 Training Algorithm of Co RA |
| Open Source Code | No | The paper does not include an unambiguous statement about releasing code or a direct link to a source-code repository for the described methodology. |
| Open Datasets | Yes | We use seven datasets including KDD99, MNIST, Caltech-256, and Image Net etc. ... MNIST. It has 70,000 training and test samples from 10 digit classes. ... Fashion MNIST. The dataset composes of a training set of 60,000 examples and a test set of 10,000 examples. ... Image Net-20, This dataset consists of images in 20 semantic concepts from Image Net dataset... Caltech-101. This dataset consists of 101 classes of images. ... CIFAR-10. There are 50,000 training and 10,000 test images from 10 categories. ... Caltech-256. This dataset contains 256 object classes with a total of 30,607 images. |
| Dataset Splits | No | The paper describes how training and test data are used (e.g., "Positive Training Data" and "Unlabeled Test Data" in Fig. 2), but it does not specify a separate validation dataset split with explicit percentages, counts, or a detailed methodology for splitting beyond the train/test distinction in their semi-supervised setting. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., GPU/CPU models, memory amounts, or cloud instance types) used for running its experiments. |
| Software Dependencies | No | The paper mentions using a "pre-trained Vgg-16 network" and "Re LU as the activation function" but does not specify any software names with version numbers for reproducibility (e.g., Python, TensorFlow, PyTorch versions). |
| Experiment Setup | Yes | We tuned λ for different tasks and observed that λ = 0.1 produces promising results for all our experiments. To optimize this loss function, stochastic gradient descent (SGD) method is adopted for training our model. ... In our experiments, the architecture of the encoder is [784, 64, 32], the decoders use a symmetric architecture. ... We use Re LU as the activation function of the hidden layers. |